DNAshape: a method for the high-throughput prediction of DNA structural features on a genomic scale.

Zhou T, Yang L, Lu Y, Dror I, Dantas Machado AC, Ghane T, Di Felice R, Rohs R. · 2013

Read DNAshape: a method for the high-throughput prediction of DNA structural features on a genomic scale. by Zhou T, Yang L, Lu Y, Dror I, Dantas Machado AC, Ghane T, Di Felice R, Rohs R. (2013) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc3692085.txt.

Introduction

Zhou and colleagues present DNAshape, a method and web server that predicts the three-dimensional shape of DNA from sequence alone, at genome scale. Proteins read DNA partly through its shape, features like groove width and twisting, but measuring shape experimentally for millions of sites is impossible.

Their method slides a window along the sequence and looks up shape values mined from Monte Carlo simulations, returning minor groove width, roll, propeller twist, and helix twist almost instantly. Predictions held up against crystal and NMR structures, cleavage data, and molecular dynamics checks. The free server is meant to let genome researchers add shape to the sequence analyses they already run.

Catalog record

CREATORZhou T, Yang L, Lu Y, Dror I, Dantas Machado AC, Ghane T, Di Felice R, Rohs R.
YEAR2013
COLLECTIONPAPERS
FORMATTXT
FILE SIZE30 KB
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC3692085

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

DNAshape: a method for the high-throughput prediction of DNA structural features on a genomic scale By Tianyin Zhou, Lin Yang, Yan Lu, Iris Dror, Ana Carolina Dantas Machado, Tahereh Ghane, Rosa Di Felice, Remo Rohs Published in: Nucleic Acids Research DOI: 10.1093/nar/gkt437 ABSTRACT We present a method and web server for predicting DNA structural features in a high-throughput (HT) manner for massive sequence data. This approach provides the framework for the integration of DNA sequence and shape analyses in genome-wide studies. The HT methodology uses a sliding-window approach to mine DNA structural information obtained from Monte Carlo simulations. It requires only nucleotide sequence as input and instantly predicts multiple structural features of DNA (minor groove width, roll, propeller twist and helix twist). The results of rigorous validations of the HT predictions based on DNA structures solved by X-ray crystallography and NMR spectroscopy, hydroxyl radical cleavage data, statistical analysis and cross-validation, and molecular dynamics simulations provide strong confidence in this approach. The DNAshape web server is freely available at http://rohslab.cmb.usc.edu/DNAshape/. INTRODUCTION An increasing number of structural biology and genomics studies associate protein–DNA binding with the recognition of the three-dimensional DNA structure, or ‘DNA shape’ (1). DNA shape readout (2) plays an important role in determining the DNA binding preferences of transcription factors (3,4) and other DNA binding proteins (5–8). Whereas DNA shape is sequence dependent, degeneracy in the sequence–structure relationship enables the formation of very similar shapes by dissimilar sequences or, in turn, dramatic effects on structure in an extended region by only a single-nucleotide substitution (9). In early studies, all of the available crystal structures of DNA fragments and protein–DNA complexes were analysed to derive the average conformations for the 10 unique dinucleotides (10). Based on the many more high-resolution structures that have been solved and analysed in recent years, it is now apparent that longer DNA segments must be characterized to capture the sequence–structure degeneracy of DNA (1). Such structural information, which can be retrieved from X-ray crystallography or NMR spectroscopy data, ideally provides information on the three-dimensional structure of a DNA binding site prior to and after protein binding. However, such data are unavailable for most sequences for their unbound or ‘naked’ states (11). Consequently, molecular simulations are the only available approach to deduce atomic information on intrinsic DNA structure. Recent efforts to characterize the structures of all 136 unique tetranucleotides have used all-atom molecular dynamics (MD) simulations of either 136 dodecamers (12) or 39 duplexes of 18 base pairs (bp) in length (13). However, in both designs most tetranucleotides occur only in the context of a single sequence, which limits the ability for a statistically robust comparison of the simulation results with experimental data. Due to their relatively time-consuming nature, simulation methods make structural analysis of massive DNA sequence data unfeasible (see Supplementary Materials and Methods). We previously improved the efficiency of conformational sampling by reducing the number of degrees of freedom in the system. Our Monte Carlo (MC) approach uses collective and internal variables for each nucleotide (14,15), an analytic chain closure with associated Jacobians (16) and an implicit solvent description (17) to realize improved sampling efficiency. This approach has enabled us to study a large number of DNA fragments. Although we successfully applied our MC approach in various studies of protein–DNA recognition (3,7,11,18,19), to bring this to the genomic scale, we have recently developed the methodology for facilitating MC data in high-throughput (HT) studies of DNA shape (4,8,18). Through the mining of atomic resolution data obtained from MC simulations (14–16) of numerous DNA fragments, our HT approach is designed to predict various important structural features of DNA for essentially any length or number of sequences. We applied a prior version of the underlying method to the minor groove width (MGW) prediction of Hox-binding sites derived from HT in vitro assays (18). Here, we present a method that improves the accuracy of this approach and expands the prediction to additional DNA structural features. Given their importance in DNA shape readout, the predicted features include MGW, Roll, propeller twist (ProT) and helix twist (HelT) (see Supplementary Figure S1 for schematic representations) (3,14,18). While we emphasize that protein–DNA recognition is a dynamic process, the predicted structural features reveal preferred conformations that are intrinsic to a given DNA sequence. We demonstrate the robustness of our method through extensive validation with massive experimental and computational data. Our HT method underlying the DNAshape web server can be used to predict DNA structural features of the entire yeast genome at nucleotide resolution in less than 1 min on a single processor. METHODOLOGY DNA shape features at a single-nucleotide position are determined by the sequence context of the corresponding bp. The context can include the immediate neighbors of a bp or a larger number of adjacent bp. Given that pentamers account for the nearest and next-nearest neighbors of their central positions, a pentamer model is a reasonably simplified approach that takes sequence context into account. Thus, the structural features at each bp position can be characterized as a function of its pentameric environment. Assuming that the structure of each of the 512 unique pentamers is known, we can use a ‘sliding pentamer model’ to derive the structural features of DNA molecules of any length and with any sequence in an instant manner (Figure 1A). Figure 1.Pentamer model for HT prediction and validation of the HT approach using Fis-binding sites. (A) MC predictions were mined with a sliding-pentamer window, and a query table of average structural features characterizing either the central bp (e.g., MGW) or the two central bp steps (e.g., Roll) of a pentamer was assembled to predict the structural features of any length of DNA. (B) HT predictions of the average MGW over the five central bp of seven Fis-binding sites (6) correlate with the logarithm of binding affinity (red). The predictive power further increases for six sequences (blue) after one sequence with a TpA step is removed (F25; Supplementary Table S2). The MGW as a function of sequence is predicted for (C) high-affinity (Kd = 0.2 nM) and (D) low-affinity (Kd = 140 nM) binding sites using the HT (blue) and MC (red) approaches and compared with X-ray data (green) of protein–DNA complexes (PDB IDs in Supplementary Materials and Methods). The large positive MGW values observed in X-ray data are usually due to crystal packing and are not observed in solution. Spearman’s rank correlation coefficients (ρ) demonstrate the statistical similarity between the predicted and experimental MGW profiles. To assemble the pentamer data, we generated a large training dataset of MC trajectories for 2121 different DNA fragments of 12–27 bp in length (see Supplementary Table S1 for list of sequences and Supplementary Materials and Methods for further details on MC method). Our MC data provide full coverage of all 512 unique pentamers, each occurring on average 44 times in the training dataset. The multiple occurrence of each pentamer in different sequence contexts is used to account for the effect of the flanking sequences through averaging. For each MC prediction, the structural features at single-nucleotide resolution were analysed with Curves (20). We used a modified definition for assigning MGW measurements to a bp, in which the MGW definition is independent of the usage

Read the complete text